How the public edition works
The application, ONNX Runtime assets, and chunked models are served as static files. Inference runs in the browser on the user's own device, using WebGPU when suitable or WebAssembly on the CPU.
No account or private backend
The free edition has no phone login, OTP, subscription, checkout, administration panel, database, or image-upload inference endpoint.
Human verification remains essential
BeshmarAI provides a suggested count. It does not replace pharmacy procedures, professional judgment, legal obligations, or a final human check.
Open public source
The public source is separated from the private platform and can be reviewed on GitHub. See the safety guide before operational use.